Thermal images to predict the thermal comfort index for Girolando heifers in the Brazilian semiarid region

• DOI: 10.1016/j.livsci.2021.104667
• Publication Date: 2021-08-09T05:44:35Z
• Authors: Pedro Henrique Di...; Gledson Luiz Pont...; Héliton Pandorfi; Marcos Vinícius d...; Rodes Angelo Bati...; Jhon Lennon Bezer...; Taize Cavalcante ...; Joez André de Mor...
• Keywords: 0402 animal and dairy science; 04 agricultural and veterinary sciences

Abstract The objective of this research was to use image analysis techniques to extract characteristics of interest and to establish a model of animal thermal comfort, based on the classification of thermal images of 7/8 Holstein-Gir heifers, through geostatistics and multivariate analysis. The thermal images were recorded by an infrared thermographic camera, which allowed images of the complete lateral faces of three heifers on pasture to be obtained. The analysis of the images consisted of the use of an extractor algorithm of notable points, referring to the regions of the head, neck, trunk, and limbs, to extract the matrices of the surface temperature of the animals. The temporal variability of the data was subjected to descriptive statistical analysis and spatial dependence, through geostatistics. The exploratory analysis of the data consisted of the application of the principal component analysis, in which 5 variables were admitted, head temperature (Thea, °C), cannon temperature (Tcan, °C), back temperature (Tbac, °C), udder temperature (Tudd, °C) and skin temperature (Tski, °C) of the animal, which allowed the identification of predictor variables correlated to animal comfort. The extraction of characteristics from the thermal images (06:00 am and 12:00 pm), using the ORB detector (Oriented fast and Rotated Brief), allowed to verification differences in the intensity of pixels in the images. The Gaussian model was the one that best fitted the studied variables and had a strong degree of spatial dependence (R2 > 0.900). The temperature of the head surface showed a greater variation (6 °C) between the analyzed times. PCA indicated a better correlation for Tcan, Tbac, Thea, and Tski in characterizing the thermal comfort of animals. The developed model proved to be efficient in characterizing the thermal comfort of heifers in the semiarid region, by determining the animal's surface temperature, with a smaller number of predictor variables.